Proceedings of The Physiological Society

University College London December 2005 (2006) Proc Physiol Soc 1, PC7

Poster Communications

Dynamic multivariate Granger causality analysis of neural and muscular signals

Wang, Shouyan; Stein, John; Aziz, Tipu; Liu, Xuguang;

1. University Laboratory of Physiology, University of Oxford, Oxford, United Kingdom. 2. Nuffield Department of Surgery, University of Oxford, Oxford, United Kingdom. 3. Division of Neuroscience and Mental Health, Imperial College London, London, United Kingdom.


Analysis of functional coupling between muscular and oscillatory neural activity at different levels of the motor system has provided important insights into neural control of movement. One widely used method of estimating the functional coupling between two oscillatory signals is coherence estimation. However, this does not provide directional information, preventing further dissection of their relationships. Granger causality analysis has proved useful in revealing causal relationships between signals. In a recent study, it was performed for all pairwise combinations of oscillatory field potential activity in the beta (1430 Hz) frequency range among sensorimotor cortical recording sites in monkeys (Brovelli, 2004). We have developed autoregressive (AR) models followed by Granger causality analysis to quantify coupling of five varieties including pairwise and group interactions with or without conditional variables. The AR models were optimised and validated according to Akaikes information criteria and prediction error for individual signal pairs. Their rescaling property and robustness to noise were investigated using simulated signals. Tremor-related EMGs and local field potentials (LFPs) of the subthalamic nucleus (STN) were simultaneously recorded in Parkinsonian patients undergoing deep brain stimulation surgery. We have therefore used conditional Granger causality analysis to quantify the directional interdependence between the STN LFPs and the envelope signal of the surface EMGs related to Parkinsonian tremor. Our results from four patients showed that during persistent tremor, there was a directional causality predominantly from surface EMGs to contralateral STN LFPs corresponding to significant coherence between them at the tremor frequency. By using an adaptive Granger causality model based on autoregressive analysis with a running window, we were able to show that the LFP-EMG interdependence was bi-directional and varied with time of tremor occurrences. We conclude that the tremor-related functional correlation appeared between the STN and the contralateral muscle. This interdependence is dynamic, bi-directional, and dependent on the tremor status.

Where applicable, experiments conform with Society ethical requirements